jadhaddad92 / covid-mask-detector Goto Github PK

Great work on Pytorch lightning, but I cannot train it on multi GPU(I set gpu=2), since Panda cannot support

OS: Windows 7
GPU: Cuda 10.1
On MaskDataset

input file isn't tensor.

thanks

Originally posted by @xsatria in #3 (comment)

Describe the bug
When I run data_preparation.py on my console whit python data_preparation.py comand it start to work, but when is unzip the folder non_mask_photos something happend that stop to do it and show on the console the error: view on the screenshot
To Reproduce
Steps to reproduce the behavior:

1. Go to cmd
2. type python data_preparation.py
3. wait a few minutes
4. see the error

Expected behavior
My expected behavior was that the code unzip all the data complete and show me a doc whit the dataset whit the id (0 or 1) (mask or not)

Screenshots

Desktop (please complete the following information):

• OS: Windows 10 pro 1909
• python 3.8.1

Describe the bug
Running video.py with the provided model ('face_mask.ckpt') with any video, fails to load the model state_dict with KeyError: 'state_dict'

To Reproduce
Steps to reproduce the behavior:
Download repo and create a virtualenv(in PyCharm) from requirements.txt.
PyTorch, Pandas and PyTorch Lightning do not install out of the box, they need to be installed manually because of version errors.

Expected behavior
Output a video with box overlays over faces with or without masks.

Screenshots

Desktop (please complete the following information):

• OS: Windows 10
• Python Version 3.7.7
• PyTorch 1.5 last stable version

Good Job with high accuracy for single person, but the accuracy lowers down for videos on the internet. I tested it with the video of Trump and his partners wearing masks. It detects very few faces and only in few frames. You can see the result Here. I also used a high resolution version of the video but nothing changed. What is the problem and how to resolve?